An "Oddity" in Significance Testing
In hypothesis testing, the alternative hypothesis is an important "structural component." In fact, it's just as important as the null hypothesis. That's because the researcher will consider Ha to represent the true state of affairs whenever Ho is rejected. Before any data are collected and any conclusion is reached, therefore, the null and alternative hypotheses have "equal standing" within the hypothesis testing procedure.
In contrast, there is no alternative hypothesis in significance testing. The only hypothesis that's involved in this form of inferential statistics is the null hypothesis. After the data are collected and analyzed, the researcher does not "turn to Ha" (if Ho seems implausible) because there is no alternative hypothesis to which one can turn. Instead, the researcher simply presents the data-based p-level as an indication of how unlikely it is that the sample came from a population situation wherein Ho is true.
Paradoxically, the p-value in significance requires the researcher decide whether to conduct a one-tailed or a two-tailed test. And as you learned in Chapter 7, the choice between doing things in a one-tailed or two-tailed manner is made when the researcher sets up Ha to be directional or nondirectional.
Thus, significance testing does not have an alternative hypothesis "on the surface." But there is one lurking in the background. It's not something that plays a role when conclusions are reached following the data analysis; it does, however, play an important role during the data analysis.
Copyright © 2012
Schuyler W. Huck